When explosive AI growth collides with systemic supply bottlenecks, Morgan Stanley states that the market's current optimism about this AI revolution may still be seriously underestimating its true explosive power and depth.
On April 11, according to Chase Trading Desk, the core judgment of Morgan Stanley's latest research report is: Top large language models (LLMs) are undergoing a &#34;non-linear capability leap,&#34; and computing power demand has already shown a trend of systemically exceeding supply.
From early January to March 2026, global weekly token usage surged from 6.4 trillion to 22.7 trillion, an increase of about 250% in just three months. Some LLM service providers have been forced to implement usage caps for their users. Morgan Stanley predicts that the future growth rate of computing power demand will be approximately 3 times the CAGR forecast for NVIDIA's computing power supply, and the computing power shortage will persist and continue to intensify long-term.
Energy is another &#34;ticking time bomb.&#34; Morgan Stanley's model predicts that U.S. data centers will face an electricity deficit of approximately 55 gigawatts between 2025 and 2028. Previously, $18 billion in data center projects were directly canceled, and another $46 billion in projects were postponed. Even with various &#34;rapid power supply&#34; solutions such as combined natural gas turbines, fuel cells, and Bitcoin site conversions, the net electricity gap could still reach 18% to 30% of the total U.S. data center deployment scale during the same period.
The impact of AI on the labor market has begun to appear. Morgan Stanley's survey shows that in the five industries most affected by AI, AI has led to 11% of positions being eliminated in the past 12 months, and another 12% of job openings were not refilled; new hires accounted for only 18%, resulting in a net layoff rate of approximately 4%. The report estimates that 90% of occupations will be affected to some extent by AI automation or augmentation.
<h2>Large Model Capabilities &#34;Step-Change Mutation&#34;: The Situation is More Extreme Than Market Expectations</h2>
Morgan Stanley's report lists the &#34;non-linear leap in frontier large model capabilities&#34; as one of the most important theme drivers for 2026, citing a large amount of data to support its judgment that &#34;the situation is far more extreme than market expectations.&#34;
According to the latest analysis from third-party firm METR, the most advanced large models can now independently complete complex tasks for over 15 hours—whereas, based on existing scaling laws, the current level should be around 8 hours. Actual capabilities have significantly outpaced theoretical expectations.
Multiple independent data points collectively confirm this trend:
<ul>
<li>
Continuous tracking indicators from Artificial Analysis show that AI capabilities are still advancing rapidly;
</li>
<li>
OpenAI CEO Sam Altman publicly warned at the India AI Summit: &#34; The world is not ready, extremely capable models are coming&#34;;
</li>
<li>
Researchers have used DNA sequencing and DeepMind's AlphaFold tool to develop cancer vaccines for their pet dogs;
</li>
<li>
A reader experiment by New York Times tech editor Kevin Roose showed that 54% of readers preferred AI-generated articles over human writing;
</li>
<li>
Frontier LLMs have demonstrated the ability to solve open problems in physics;
</li>
<li>
Reports suggest that an unreleased model represents a &#34;step-change mutation in capability&#34; in areas such as software programming, academic reasoning, and cybersecurity.
</li>
</ul>
The report also quotes a prediction from Leo Aschenbrenner's paper &#34;Context-Awareness&#34;:
<blockquote>
&#34;The possibility of achieving AGI (Artificial General Intelligence) by 2027 is surprisingly high. In the four years from GPT-2 to GPT-4, we leaped from a preschool level to a smart high school student level... If we take the same intellectual leap again, where will it take us? Likely to models capable of surpassing PhDs and top experts in all professional fields.&#34;
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/fb901ba3-970a-49b3-a604-cc03a31912fb.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/fb901ba3-970a-49b3-a604-cc03a31912fb.png"/>
</blockquote>
<h2>Huge Gap in Computing Power Supply and Demand: 250% Token Growth Driven by a 3x Demand Discrepancy</h2>
If the leap in large model capabilities is the &#34;engine on the demand side,&#34; the severe inadequacy of computing power supply is the &#34;ceiling on the supply side.&#34; Morgan Stanley identifies &#34;computing power demand systematically exceeding supply&#34; as the most critical market theme for 2026.
The report states the data is very intuitive:
<ul>
<li>
According to actual tracking data from the OpenRouter platform, from early January to March 2026, global weekly token usage increased from 6.4 trillion to 22.7 trillion, an increase of approximately 250% in three months;
</li>
<li>
The rapid adoption of agent AI tools (represented by OpenClaw) has significantly accelerated the demand-side explosion;
</li>
<li>
Multiple LLM service providers have begun to set token usage limits for users to cope with runaway demand growth;
</li>
<li>
Morgan Stanley predicts that the overall growth rate of computing power demand will be approximately 3 times the CAGR forecast for NVIDIA's computing power supply;
</li>
<li>
Three concurrent driving forces are compounding demand: continuous expansion of AI use cases, non-linear increase in AI task complexity, and accelerated broadening of AI adoption.
</li>
</ul>
At the application level, software programming is currently the single largest Token consumption use case for all LLMs, and this field is dominated by proprietary (closed-source) models.
Morgan Stanley's &#34;Intelligence Factory&#34; model reveals another key logic: as the chip generation transitions from Blackwell to Rubin GPUs, the average Token price is expected to decrease by over 70%—the rapid decline in AI usage costs will further stimulate demand, creating a positive feedback loop of self-reinforcing demand.
Taking a specific example: a data center with a scale of approximately 250 megawatts, using Blackwell GPUs, with electricity costs of $100 per megawatt-hour, and running GPT-4o queries, can generate about 60% profit margin for top large model developers.
Morgan Stanley expects that actual computing power demand will reach about 3 times the previously modeled predictions. In this context, any company that can break the computing power expansion bottleneck will benefit significantly. This includes not only the chip manufacturing supply chain but also memory, optical network equipment, and core data center components. Morgan Stanley is extremely optimistic about these &#34;Merchants of Compute,&#34; believing they will directly benefit from this systemic supply-demand imbalance.
<h2>Energy is the Lifeline of AI: The 55 Gigawatt Deficit and the Race for &#34;Off-Grid&#34; Solutions</h2>
Electricity has become the most critical physical constraint for AI infrastructure expansion. Morgan Stanley's &#34;AI Power&#34; in-depth analysis model has reached alarming conclusions.
Between 2025 and 2028, U.S. data center developers will face an electricity supply gap of approximately 55 gigawatts. Simultaneously, $18 billion in data center projects have been directly canceled due to community opposition and concerns over electricity prices, while another $46 billion in projects have been postponed. Multiple obstacles constraining data center growth are escalating concurrently: competition for grid access resources, power equipment shortages, labor deficits, and local political resistance.
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/62c61ec4-70f5-43c7-97e5-ed1d7401dc06.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/62c61ec4-70f5-43c7-97e5-ed1d7401dc06.png"/>
Facing this gap, Morgan Stanley has outlined four categories of &#34;Time to Power&#34; solutions:
<blockquote>
<ul>
<li>Natural Gas Turbines: Can alleviate a 15-20 GW deficit, 90% probability of success;</li>
<li>Bloom Energy Fuel Cells: Can alleviate a 5-8 GW deficit, 90% probability of success;</li>
<li>Deploying Data Centers at Existing Nuclear Power Plants: Can alleviate a 5-15 GW deficit, 75% probability of success;</li>
<li>Converting Bitcoin Mining Farms to Data Centers: Can alleviate a 10-15 GW deficit, 90% probability of success.</li>
</ul>
</blockquote>
However, even when the weighted contribution of all these solutions is aggregated, Morgan Stanley's baseline calculation shows that the net electricity deficit by 2028 will still be equivalent to 18% to 30% of the total U.S. data center deployment scale during the same period.
From a strategic perspective, Meta has begun to take proactive action—providing funding for TerraPower's commercialized sodium-cooled fast reactor project and directly investing in Louisiana's power infrastructure.
Morgan Stanley believes this may be a strategic signal that AI giants are beginning to systematically control energy infrastructure to secure their computing power lifeline.
<h2>Initial Signs of Employment Impact, AI Adoption Economic Value Exceeds S&amp;P 500 Pre-Tax Profits by 25%</h2>
Morgan Stanley's survey data and model calculations reveal the early and profound impact of AI on the labor market.
In the five industries most significantly affected by AI (consumer retail, real estate management and development, transportation, medical equipment and services, automotive and parts), Morgan Stanley's field survey shows:
<blockquote>
<ul>
<li>In the past 12 months, AI has led to 11% of positions being directly eliminated;</li>
<li>Another 12% of job openings were not refilled;</li>
<li>New hires accounted for 18%, resulting in a net layoff rate of approximately 4%;</li>
<li>Notably, smaller businesses showed the weakest new hiring – this may reflect their more agile and rapid adoption of AI.</li>
</ul>
</blockquote>
From a macro perspective, Morgan Stanley estimates that 90% of occupations will be affected to some extent by AI automation or augmentation, typically not by &#34;eliminating entire job roles&#34; but by &#34;reconfiguring the task structure within roles.&#34;
In terms of quantitative economic value, the Total Addressable Market (TAM) for AI adoption calculated by Morgan Stanley is equally astonishing:
<blockquote>
<ol>
<li>The value TAM corresponding to the cost reduction potential brought by &#34;AI automation&#34; exceeds 25% of the S&amp;P 500's expected adjusted pre-tax profits for 2026;
<img src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/4d95df85-8ff5-4671-8c25-ab2ee3532510.png?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://wpimg-wscn.awtmt.com/4d95df85-8ff5-4671-8c25-ab2ee3532510.png"/></li>
<li>This &#34;AI automation&#34; cost reduction is equivalent to over 40% of the total employee compensation expenses;</li>
<li>Among these, agent AI (software level) and embodied AI (physical level, represented by robots) contribute nearly half each;</li>
<li>By industry, the economic potential of AI adoption relative to pre-tax profits is highest in consumer retail, real estate management, transportation, and medical equipment.</li>
</ol>
</blockquote>
<h2>The &#34;Moat&#34; Disrupted by AI: What Assets Will Truly Hold Value in the Age of AI?</h2>
As AI capabilities accelerate their leap, a core question becomes increasingly urgent: In a world where AI can replicate almost anything, what assets possess true defensibility? Morgan Stanley's report quotes investor Michael Bloch's framework to propose a key distinction:
<blockquote>
&#34; AI compresses the time required to get things done, but it cannot compress the time required for things to happen naturally. This distinction is the most important screening criterion in current investment.&#34;
</blockquote>
Based on this, the types of assets with truly defensive moats include five categories:
<ol>
<li>Continuously accumulating proprietary data—not static datasets, but dynamic data generated through defensible business operations;</li>
<li>Network effects—each new user makes the product more valuable to other users; network advantages that have accumulated liquidity will become more prominent as AI lowers the barrier to creating competitors;</li>
<li>Regulatory licenses—bank licenses take years, FDA approvals take years; regulatory barriers expand with AI capabilities, not shrink;</li>
<li>Large-scale capital deployment capability—when bottlenecks shift from software to physical infrastructure, the ability to mobilize massive capital itself becomes a core advantage of the era;</li>
<li>Physical infrastructure—factories, power plants, data centers... the laws of physics set an unbreakable lower bound for time, and the first-mover advantage expands every month.</li>
</ol>
The report further lists eight categories of assets that may appreciate in the &#34;transformative AI&#34; era, covering: real estate with physical scarcity (AI infrastructure land, industrial properties), AI application adopters with pricing power, luxury goods and unique services, platforms with network effects, authentic and unique human experiences, regulatory franchises, proprietary data and brands, and critical semiconductor assets (advanced process chip fabs, ASML's EUV lithography machines, rare earth processing capabilities).
<hr/>

NVDU

NVDL

NVDS

Morgan Stanley believes that the market's current optimism about this AI revolution may still be seriously underestimating its true explosive power and depth. Top large models are undergoing "non-linear capability leaps," with global token usage surging 250% in three months; the growth rate of computing power demand is three times that of NVIDIA's supply growth, and the supply-demand gap will widen long-term. In terms of energy, U.S. data centers face a 55-gigawatt power deficit between 2025-2028

- Morgan Stanley's latest report highlights a significant demand for AI capabilities, indicating a 250% increase in global token usage from January to March 2026, while supply is lagging considerably.  
- The report forecasts a continuing power deficit of around 55 gigawatts for data centers in the U.S. between 2025 and 2028, as project cancellations and delays arise due to rising energy costs.  
- AI's impact on the labor market is evident, with 11% of jobs cut in the past year across affected sectors, while 90% of roles may experience some level of automation or enhancement due to AI.

Wallstreetcn

Direxion Daily NVDA Bull 2X Shares

GraniteShares 2x Long NVDA Daily ETF

AXS 1.5X NVDA Bear Daily ETF

iShares Semiconductor ETF

iShares Expanded Tech Software Sector ETF

T-Rex 2X Inverse NVIDIA Daily Target ETF

Direxion Semicon Bull 3X

VanEck Semiconductor ETF

SPDR S&P Software

NVIDIA

Spdr Select Tech

YieldMax NVDA Option Income Strategy ETF

Direxion Daily NVDA Bear 1X ETF

T-Rex 2X Long NVIDIA Daily Target ETF

Top Large Models Show "Capability Mutation," Computing Power Demand "Systemically Exceeds Supply" – Morgan Stanley: "Market Optimism May Not Be Enough"